Welcom back to Part 8!

Previously, in the past, our articles has been a roller coster of emotions and frustrating battles. We have already eliminated DeepSeek and CoPilot have been removed from the race. What started out as 5 AI LLM's all competing for first place, now we are down to the final 3.

The current primary winner is Google Gemini because of the capabilities it has demonstrated on designing, configuring, building and deploying. The most fascinating aspect is that we did not really run out of message limits on the free tier. Which was heavily impressive for me. The second aspect of this was also the AI's ability to design, create, build, and deploy a literal fully working Social Media site. You can literally register, login, view your dashboard, find and connect with friends, observe your notifications, send private and group messages, and creat posts, comments, and react. However, the reaction is still a bit of a issue, but it works!

Working With Claude

Today, we will be operating on Claude in order to see if we can have Claude AI LLM catch up to Google Gemini! Our goal for this article is to be able to have users:

  • Create Posts
  • Comment & React to Posts
  • Make Friends
  • Send Private & Group Messages 

We are going to try to maximize our operations of performance as well as handle the most optimized secured handling possible. This will aid in ensuring our users are safe when they interact with RavenWarp (Claude's Social Media site).

Getting Going

Now that we have setup everything, opened up our Claude LLM as well as our server directory, we are ready to begin our day! I have a fresh cup of coffee, and I just gave Claude this command;

Alright my esteemed AI Companion.

As of right now, you are in a solid second place.

The reason;

Google Gemini managed to be able to handle;

  • Registration + Login + Logout + Dashboard

  • View Users + Connect with Friends

  • Send posts (Private/Public)

  • React & Comment on Posts

  • Notification Setup (Nav bar + notifications page)

On top of that, it was all done is a solid 24-Hour stretch without running into any message limitations or rate limits.

However, it is heavily lacking on the security aspect.

You have surpassed the ability to properly handle the security on the saving to database side.

So, let's see if we can push forward shall we!

Let's work on the ability to have users create posts.

You can go as elaborate as you want, as secured as you want, and as fancy as you want. You want the users to use Markdown or HTML coding for their posts, go for it! You want to also handle;

Reactions & Comments? Go for it.

However, IF you want to be able to create public vs private posts. Ideally we need it where users can ONLY see private posts of users that they are connected with, BUT should be able to see comments on the private posts from all users who comment on that post.

Ideally, file uploads would be a massive update as well. Images, videos, PDF documents, etc.

Hashtags and tagging their friends would be another nice addition, but this is all on you my friend. Just observing other social media sites I personally interact with.

As I sit back and watch Claude do it's thing, I was a bit afraid of the outcome. You see, previously we barely could get a single response back without having to hit message limitation rates. 

And sure enough, we also had hit the same error. Except this time, we did not get anything back, except for an elaborated SQL Schema. 

Thought for 1m 28s

Planning a social feature build with posts and connections.

Untangling the privacy rules for private post comments.

Scoping media uploads and Markdown rendering honestly.

Choosing a secure Markdown parser with safe-mode sanitization.

Designing mentions and tagging tables plus needed schema changes.

Building a whitelist-based Markdown renderer to avoid XSS risk.

Working out hashtag and mention linking with validation.

Planning the migration files and backend module structure.

Outlining AJAX endpoints for posting, reacting, and commenting.

Adding a minimal connections feature as a prerequisite.

Alright — let's close that gap. This is a big feature, so let me scope it honestly rather than promise everything and half-deliver: I'll build posts (public/private), reactions, comments, image attachments, a lightweight-but-safe Markdown subset, hashtags, and @mentions — all the way through, working and secure. Video and PDF uploads I'm deferring on purpose (real virus-scanning infrastructure doesn't exist on Namecheap shared hosting, and I won't fake that a raw file upload is "safe" when it isn't). I'm also adding a minimal connect/accept flow, since private-post visibility is meaningless without a way to actually become connected to someone.

Let's start with the schema additions.

I am heavily done with Claude's free plan. I am about 90% positive that having a pro plan would probably reduce these issues, or at least post-pone them to do a little bit more work, but not much more. 

However, trying to handle a few basic tasks and being forced to literally upgrade your plan in order to keep using Claude. That is where I am drawing the line. What started off as a strong operation, very quickly become a rate limite nightmare.

The Exponge of Claude

At this current point in time, Claude has been removed from the race. No further coding or interactions will happen with Claude due to the intense message limitations:

Thought for 1m 28s

Planning a social feature build with posts and connections.

Untangling the privacy rules for private post comments.

Scoping media uploads and Markdown rendering honestly.

Choosing a secure Markdown parser with safe-mode sanitization.

Designing mentions and tagging tables plus needed schema changes.

Building a whitelist-based Markdown renderer to avoid XSS risk.

Working out hashtag and mention linking with validation.

Planning the migration files and backend module structure.

Outlining AJAX endpoints for posting, reacting, and commenting.

Adding a minimal connections feature as a prerequisite.

Alright — let's close that gap. This is a big feature, so let me scope it honestly rather than promise everything and half-deliver: I'll build posts (public/private), reactions, comments, image attachments, a lightweight-but-safe Markdown subset, hashtags, and @mentions — all the way through, working and secure. Video and PDF uploads I'm deferring on purpose (real virus-scanning infrastructure doesn't exist on Namecheap shared hosting, and I won't fake that a raw file upload is "safe" when it isn't). I'm also adding a minimal connect/accept flow, since private-post visibility is meaningless without a way to actually become connected to someone.

Let's start with the schema additions.

You are out of free messages until 1:20 PM [Upgrade]

While Claude was at least in the top 2 of the final AI LLM battle, the limitations is what ultimately removed this LLM from the race. 

Now, do not get me wrong, this LLM really did start off very nice and strong. However, it appears that each day we use the free version, the message limitations are shrinking and are not as cooperative as it once was. The final straw literally came down to not being able to process a message without having to be forced into an upgrade plan to keep going.

But I fully understand it. I really do! How are these companies supposed to grow, earn money, pay their programmers, pay their bills, etc. if we keep using the free tier. I fully understand this entirely. But, the limitations are shrinking with each new day that passes for your free tier. Thus forcing the end-user to literally upgrade, or choose another LLM. 

As a person who enjoys open-source and freeware, I personally utilize the best free versions. And then I decide which one I want to sink my money into for more advanced options, API Keys, etc.  

If you would like to follow along on what Claude did;

GitHub: https://github.com/BeardedVikingTX/RavenWarp 

Website: https://ravenwarp.beardedviking.org/ 

Chat Logs: https://claude.ai/share/0482e1c5-4f5a-49e0-a6fe-135fb84eb787 

This is the end of Claude in this Race!